Introduction

1 2 3 4 5 6

Key Questions

  • How does HIV persist and migrate across CNS and peripheral tissues despite systemic viral suppression?

Overview

This presentation summarizes:

  • Viral phylogenetics & BEAST modeling
  • CNS reservoir compartmentalization
  • Migration dynamics across tissues & BBB



Cohort Workflow

Last Gift Cohort

Body Donation

Autopsy

Tissue Collection

HIV Sequencing

Phylogenetics and Modeling

OBJECTIVES

Blood Brain Interface

Image attribution: BioRender.
(click anywhere to close)

Objectives

Aim 1

  • Primary objectives:
    • model viral migration events across the Blood Brain Barrier (CNS ↔︎ periphery).
    • identify CSF markers and reservoir characteristics associated with dynamics
  • Secondary objectives:
    • model viral migration events within the CNS.
    • identify CSF markers and reservoir characteristics associated with CNS dynamics

Aim 2

  • Identify CNS-specific clusters (reservoirs)
  • Characterize:
    • Composition
    • Timing
    • Renewal from periphery
    • Reseeding capacity into peripheral reservoirs

APPROACH

Aim 1

Data Processing

  • SGA HIV env clones across tissues
  • CSF biomarkers & clinical metadata
  • BEAST diffusion + phylogenetic modeling
  • Negative Binomial & Poisson models

Workflow

Aim 2

Data Processing

  • HIV env clones across participants & tissues
  • Cleanup & alignment
  • Phylogenetic reconstruction
  • Identify CNS-specific supported clusters
  • Bayesian skyline / skygrid diffusion models
  • Timing (tMRCA) of clusters
  • Migration flows between CNS clusters & other tissues
  • Modeling factors associated with migration

Workflow

Sampling

Note

  • Our analyses included 20 participants with sequences from both the CNS and Periphery.
  • A total of 1859 SGA clones were included from 45 distincts tissues.
  • Luminex CSF data available for 34 markers.

AIM 1: Viral Dynamics

Maximum Clade Credibility Tree

Viral Suppression: Individuals with detectable viral load >1000 copies/µl are indicated with an *.

Time-Scaled Maximum Clade Credibility Tree

Viral Suppression: Individuals with detectable viral load >1000 copies/µl are indicated with an *.

HIV Migration Networks

Viral Suppression: Individuals with detectable viral load >1000 copies/µl are indicated with an *.

HIV Flow across the BBB through time

Viral Suppression: Individuals with detectable viral load >1000 copies/µl are indicated with an *.

AIM 1: Reservoirs Dynamic and CSF Markers

Summary

Objectives

  • Primary objectives:
    • model viral migration events across the Blood Brain Barrier (CNS ↔︎ periphery).
    • identify CSF markers and reservoir characteristics associated with dynamics
  • Secondary objectives:
    • model viral migration events within the CNS.
    • identify CSF markers and reservoir characteristics associated with CNS dynamics

Approach

  • We compared a patient-level aggregated negative-binomial model (AGG-NB) with a state-level hierarchical negative-binomial model (STATE-NB) that accounted for variation across individual BEAST states.

Data Preparation

  • Aggregated (pid-level counts):We first fit an aggregated-count model (mod_agg_nb), where counts were summed across BEAST states for each participant.”
  • Longitudinal (state-level / transition-level): We then fit a state-level model (mod_long_nb), leveraging the full longitudinal BEAST reconstruction.

Exposure (log)

  • Aggregated (pid-level counts): product of n_states* n_seq_cns*n_seq_periphery.
  • Longitudinal (state-level / transition-level): product of n_seq_source*n_seq_destination

Data Input

Exposure

Conceptual motivation

In a transition model, exposure represents the “opportunity” for observing a migration event between compartments. We considered the following definition:

  • Aggregated (pid-level counts): product of n_states* n_seq_cns*n_seq_periphery.
  • Longitudinal (state-level / transition-level): product of n_seq_source*n_seq_destination

Equation

\[ n_{\text{states}, i} : \text{number of Markov states in run } i \times \text{number of sequences from CNS} \times \text{number of sequences from Periphery} \]

Model Code

data %>%
  mutate(
    pairs_cns    = ntissues_cns * (ntissues_cns - 1) / 2,
    pairs_periph = ntissues_periph * (ntissues_periph - 1) / 2,
    
    exposure = case_when(
      migration_type == "cross_BBB" & direction == "CNS to Periph" ~ 
        n_states * nseq_cns * nseq_periph,
      
      migration_type == "cross_BBB" & direction == "Periphery to CNS" ~ 
        n_states * nseq_periph * nseq_cns,
      
      migration_type == "within_CNS" ~ 
        n_states * pairs_cns * (nseq_cns^2),
      
      migration_type == "within_peripheral" ~ 
        n_states * pairs_periph * (nseq_periph^2)
    )
  )

CSF Markers

Data Preparation for Migration Models

Summaries

  • Count migration/transition events from and toward the CNS and within CNS
  • Compute number of discrete viral states per run
  • Count SGA clones sampled in CNS and in Periphery

Exposure Definition

  • Aggregated: # states × # CNS sequences × # Periphery sequences
  • long format: # sequences from × # sequences to

Filtering

  • Exclude viremic individuals
  • Exclude individuals without CSF data

Data Transformations

  • Log-transform exposure
  • Log1p-transform CSF marker values and reservoir measures

Workflow

Model Specification: Negative Binomial GLMM

Notes

  • Outcome: count of CNS-related migration/transition events
  • Offset: log(exposure)
  • Predictor: log1p-transformed CSF marker
  • Random effect for participants

Equation

\[ \log(\mu_i) = \beta_0 + \beta_1 \cdot \text{Marker}_i + b_{\text{pid}(i)} + \log(\text{Exposure}_i) \]

fit_models <- function(marker, data_in) {
  formula <- as.formula(
    paste0(
      "n_events ~ ", marker,
      " + age + sex + last_cd4_t_cell_count + duration_infection_years +
        (1 | pid) + offset(log_exposure)"
    )
  )

  glmmTMB(
    formula,
    data = data_in,
    family = nbinom2(),
    control = glmmTMBControl(
      optimizer = optim,
      optArgs = list(method = "BFGS", maxit = 5000)
    )
  )
}

fit_models <- function(marker, data_in) {
  formula <- as.formula(
    paste0(
      "n_events ~ ", marker,
      " + age + sex + last_cd4_t_cell_count + duration_infection_years +
      dna_level_from + dna_level_to +
        (1 | pid) + offset(log_exposure)"
    )
  )

  glmmTMB(
    formula,
    data = data_in,
    family = nbinom2(),
    control = glmmTMBControl(
      optimizer = optim,
      optArgs = list(method = "BFGS", maxit = 5000)
    )
  )
}

Model Results: Aggregated (pid-level counts)

Viral Suppression: Individuals with detectable viral load >1000 copies/µl are excluded (eligible: LG44 and LG48).

Associations between CSF biomarkers and counts of migration events from the CNS to the periphery
Negative Binomial Models
Marker Samples Detectable1 Effect Status exp(Beta) 95% CI p-value Significance Direction
Eotaxin (CCL11) 13 Ok ✅ 1.042 0.7996–1.36 0.759 ⬆
GM-CSF 8 Ok ✅ 0.890 0.7506–1.06 0.180 ⬇
GRO-alpha (CXCL1) 13 Ok ✅ 0.988 0.9805–1.00 0.001 ** ⬇
IL-1α 1 Sparse Data ❌ 0.236 0.0031–17.88 0.513 ⬇
IL-1β 5 Ok ✅ 0.839 0.6518–1.08 0.173 ⬇
IL-1RA 12 Ok ✅ 1.000 0.9995–1.00 0.000 *** ⬇
IL-2 2 Sparse Data ❌ 0.419 0.2598–0.67 0.000 *** ⬇
IL-5 5 Ok ✅ 0.985 0.8074–1.20 0.877 ⬇
IL-6 12 Ok ✅ 0.998 0.9971–1.00 0.000 *** ⬇
IL-7 13 Ok ✅ 0.245 0.0441–1.37 0.109 ⬇
IL-8 (CXCL8) 13 Ok ✅ 0.998 0.9954–1.00 0.089 . ⬇
IL-9 1 Sparse Data ❌ 1.165 0.6152–2.20 0.640 ⬆
IL-10 8 Ok ✅ 0.511 0.2750–0.95 0.033 * ⬇
IL-15 5 Ok ✅ 0.491 0.1756–1.38 0.176 ⬇
IL-17A/CTLA-8 1 Sparse Data ❌ 0.327 0.0442–2.42 0.274 ⬇
IL-18 13 Ok ✅ 1.005 0.9827–1.03 0.686 ⬆
IL-31 1 Sparse Data ❌ 0.926 0.6711–1.28 0.641 ⬇
IP-10 (CXCL10) 13 Ok ✅ 1.000 0.9969–1.00 0.926 ⬇
MCP-1 (CCL2) 13 Ok ✅ 1.000 0.9988–1.00 0.953 ⬇
MIP-1α (CCL3) 13 Ok ✅ 0.871 0.7684–0.99 0.031 * ⬇
MIP-1β (CCL4) 13 Ok ✅ 0.965 0.9424–0.99 0.002 ** ⬇
RANTES (CCL5) 13 Ok ✅ 1.009 0.9194–1.11 0.849 ⬆
SDF-1α 13 Ok ✅ 1.000 0.9994–1.00 0.751 ⬆
TNF-α 2 Sparse Data ❌ 0.326 0.2118–0.50 0.000 *** ⬇
1 Number of samples with ≥1 measure(s) above the LOD across the 3 replicates.
Associations between CSF biomarkers and counts of migration events from the periphery to the CNS
Negative Binomial Models
Marker Samples Detectable1 Effect Status exp(Beta) 95% CI p-value Significance Direction
Eotaxin (CCL11) 16 Ok ✅ 0.970 0.864–1.09 0.607 ⬇
GM-CSF 11 Ok ✅ 0.878 0.780–0.99 0.031 * ⬇
GRO-alpha (CXCL1) 16 Ok ✅ 0.986 0.981–0.99 0.000 *** ⬇
IL-1α 1 Sparse Data ❌ 0.230 0.014–3.92 0.310 ⬇
IL-1β 7 Ok ✅ 0.873 0.753–1.01 0.072 . ⬇
IL-1RA 15 Ok ✅ 1.000 1.000–1.00 0.001 *** ⬇
IL-2 2 Sparse Data ❌ 0.380 0.304–0.48 0.000 *** ⬇
IL-5 7 Ok ✅ 0.915 0.713–1.17 0.486 ⬇
IL-6 15 Ok ✅ 0.998 0.997–1.00 0.000 *** ⬇
IL-7 16 Ok ✅ 0.866 0.232–3.23 0.831 ⬇
IL-8 (CXCL8) 16 Ok ✅ 0.999 0.997–1.00 0.162 ⬇
IL-9 1 Sparse Data ❌ 1.205 0.712–2.04 0.487 ⬆
IL-10 10 Ok ✅ 0.916 0.765–1.10 0.339 ⬇
IL-15 7 Ok ✅ 0.378 0.183–0.78 0.009 ** ⬇
IL-17A/CTLA-8 2 Sparse Data ❌ 0.584 0.225–1.51 0.268 ⬇
IL-18 16 Ok ✅ 1.011 0.995–1.03 0.179 ⬆
IL-31 1 Sparse Data ❌ 0.908 0.718–1.15 0.422 ⬇
IP-10 (CXCL10) 16 Ok ✅ 1.000 0.999–1.00 0.398 ⬇
MCP-1 (CCL2) 16 Ok ✅ 1.000 0.999–1.00 0.862 ⬇
MIP-1α (CCL3) 16 Ok ✅ 0.904 0.839–0.97 0.008 ** ⬇
MIP-1β (CCL4) 16 Ok ✅ 0.983 0.968–1.00 0.026 * ⬇
RANTES (CCL5) 16 Ok ✅ 0.990 0.933–1.05 0.748 ⬇
SDF-1α 16 Ok ✅ 1.000 0.999–1.00 0.494 ⬇
TNF-α 3 Sparse Data ❌ 0.362 0.234–0.56 0.000 *** ⬇
1 Number of samples with ≥1 measure(s) above the LOD across the 3 replicates.
Associations between CSF biomarkers and counts of migration events across the BBB
Negative Binomial Models
Marker Samples Detectable1 Effect Status exp(Beta) 95% CI p-value Significance Direction
Eotaxin (CCL11) 16 Ok ✅ 1.014 0.8655–1.19 0.866 ⬆
GM-CSF 11 Ok ✅ 0.892 0.7872–1.01 0.071 . ⬇
GRO-alpha (CXCL1) 16 Ok ✅ 0.988 0.9817–1.00 0.001 *** ⬇
IL-1α 1 Sparse Data ❌ 0.330 0.0082–13.26 0.557 ⬇
IL-1β 7 Ok ✅ 0.875 0.7033–1.09 0.234 ⬇
IL-1RA 15 Ok ✅ 1.000 0.9995–1.00 0.001 *** ⬇
IL-2 2 Sparse Data ❌ 0.393 0.2965–0.52 0.000 *** ⬇
IL-5 7 Ok ✅ 0.926 0.7871–1.09 0.354 ⬇
IL-6 15 Ok ✅ 0.998 0.9972–1.00 0.001 *** ⬇
IL-7 16 Ok ✅ 1.010 0.2351–4.34 0.989 ⬆
IL-8 (CXCL8) 16 Ok ✅ 0.999 0.9970–1.00 0.174 ⬇
IL-9 1 Sparse Data ❌ 1.106 0.6638–1.84 0.699 ⬆
IL-10 10 Ok ✅ 0.952 0.7445–1.22 0.698 ⬇
IL-15 7 Ok ✅ 0.413 0.1850–0.92 0.031 * ⬇
IL-17A/CTLA-8 2 Sparse Data ❌ 0.825 0.2375–2.86 0.762 ⬇
IL-18 16 Ok ✅ 1.010 0.9922–1.03 0.284 ⬆
IL-31 1 Sparse Data ❌ 0.968 0.7310–1.28 0.820 ⬇
IP-10 (CXCL10) 16 Ok ✅ 1.000 0.9986–1.00 0.834 ⬆
MCP-1 (CCL2) 16 Ok ✅ 1.000 0.9989–1.00 0.854 ⬇
MIP-1α (CCL3) 16 Ok ✅ 0.876 0.7832–0.98 0.020 * ⬇
MIP-1β (CCL4) 16 Ok ✅ 0.982 0.9598–1.01 0.126 ⬇
RANTES (CCL5) 16 Ok ✅ 1.007 0.9299–1.09 0.872 ⬆
SDF-1α 16 Ok ✅ 1.000 0.9990–1.00 0.848 ⬆
TNF-α 3 Sparse Data ❌ 0.357 0.2330–0.55 0.000 *** ⬇
1 Number of samples with ≥1 measure(s) above the LOD across the 3 replicates.
Associations between CSF biomarkers and counts of migration events within CNS
Negative Binomial Models
Marker Samples Detectable1 Effect Status exp(Beta) 95% CI p-value Significance Direction
Eotaxin (CCL11) 15 Ok ✅ 1.046 8.5e-01–1.294 0.680 ⬆
GM-CSF 10 Ok ✅ 0.897 7.2e-01–1.114 0.326 ⬇
GRO-alpha (CXCL1) 15 Ok ✅ 0.987 9.8e-01–0.998 0.019 * ⬇
IL-1α 1 Sparse Data ❌ 0.001 1.5e-05–0.039 0.000 *** ⬇
IL-1β 6 Ok ✅ 0.878 6.8e-01–1.128 0.309 ⬇
IL-1RA 14 Ok ✅ 0.999 1.0e+00–1.000 0.000 *** ⬇
IL-2 2 Sparse Data ❌ 0.405 2.4e-01–0.683 0.001 *** ⬇
IL-5 7 Ok ✅ 1.127 7.6e-01–1.678 0.558 ⬆
IL-6 14 Ok ✅ 0.998 1.0e+00–0.999 0.004 ** ⬇
IL-7 15 Ok ✅ 0.602 1.3e-01–2.870 0.524 ⬇
IL-8 (CXCL8) 15 Ok ✅ 0.999 1.0e+00–1.002 0.586 ⬇
IL-9 1 Sparse Data ❌ 0.491 2.0e-01–1.225 0.127 ⬇
IL-10 9 Ok ✅ 1.019 7.2e-01–1.434 0.915 ⬆
IL-15 7 Ok ✅ 0.171 4.0e-02–0.735 0.018 * ⬇
IL-17A/CTLA-8 2 Sparse Data ❌ 1.849 2.8e-01–12.424 0.527 ⬆
IL-18 15 Ok ✅ 1.018 9.9e-01–1.047 0.212 ⬆
IL-31 1 Sparse Data ❌ 0.890 6.4e-01–1.242 0.493 ⬇
IP-10 (CXCL10) 15 Ok ✅ 1.000 1.0e+00–1.001 0.875 ⬇
MCP-1 (CCL2) 15 Ok ✅ 0.999 1.0e+00–1.001 0.281 ⬇
MIP-1α (CCL3) 15 Ok ✅ 0.915 8.0e-01–1.048 0.200 ⬇
MIP-1β (CCL4) 15 Ok ✅ 0.972 9.4e-01–1.010 0.141 ⬇
RANTES (CCL5) 15 Ok ✅ 0.920 8.2e-01–1.035 0.165 ⬇
SDF-1α 15 Ok ✅ 1.000 1.0e+00–1.001 0.537 ⬇
TNF-α 3 Sparse Data ❌ 0.462 2.0e-01–1.042 0.063 . ⬇
1 Number of samples with ≥1 measure(s) above the LOD across the 3 replicates.

Model Results: Longitudinal / state-level data

Viral Suppression: Individuals with detectable viral load >1000 copies/µl are excluded (eligible: LG44 and LG48).

Associations between CSF biomarkers and counts of migration events from the CNS to the periphery
Negative Binomial Models (per transition)
Marker Samples Detectable1 Effect Status exp(Beta) 95% CI p-value Significance Direction
Eotaxin (CCL11) 13 Ok ✅ 1.049 0.923–1.19 0.463 ⬆
GM-CSF 8 Ok ✅ 0.925 0.855–1.00 0.054 . ⬇
GRO-alpha (CXCL1) 13 Ok ✅ 0.995 0.991–1.00 0.033 * ⬇
IL-1α 1 Sparse Data ❌ 0.134 0.018–1.02 0.052 . ⬇
IL-1β 5 Ok ✅ 0.973 0.846–1.12 0.697 ⬇
IL-1RA 12 Ok ✅ 1.000 1.000–1.00 0.000 *** ⬇
IL-2 2 Sparse Data ❌ 0.695 0.532–0.91 0.008 ** ⬇
IL-5 5 Ok ✅ 0.973 0.885–1.07 0.574 ⬇
IL-6 12 Ok ✅ 0.999 0.999–1.00 0.018 * ⬇
IL-7 13 Ok ✅ 1.139 0.420–3.09 0.798 ⬆
IL-8 (CXCL8) 13 Ok ✅ 0.998 0.997–1.00 0.000 *** ⬇
IL-9 1 Sparse Data ❌ 0.853 0.591–1.23 0.394 ⬇
IL-10 8 Ok ✅ 0.608 0.455–0.81 0.001 *** ⬇
IL-15 5 Ok ✅ 0.827 0.488–1.40 0.480 ⬇
IL-17A/CTLA-8 1 Sparse Data ❌ 0.543 0.198–1.49 0.236 ⬇
IL-18 13 Ok ✅ 1.003 0.991–1.02 0.597 ⬆
IL-31 1 Sparse Data ❌ 1.023 0.863–1.21 0.795 ⬆
IP-10 (CXCL10) 13 Ok ✅ 1.000 0.998–1.00 0.877 ⬇
MCP-1 (CCL2) 13 Ok ✅ 0.999 0.999–1.00 0.000 *** ⬇
MIP-1α (CCL3) 13 Ok ✅ 0.960 0.889–1.04 0.294 ⬇
MIP-1β (CCL4) 13 Ok ✅ 0.983 0.968–1.00 0.020 * ⬇
RANTES (CCL5) 13 Ok ✅ 1.004 0.959–1.05 0.874 ⬆
SDF-1α 13 Ok ✅ 1.000 0.999–1.00 0.895 ⬇
TNF-α 2 Sparse Data ❌ 0.598 0.448–0.80 0.000 *** ⬇
1 Number of samples with ≥1 measure(s) above the LOD across the 3 replicates.
Associations between CSF biomarkers and counts of migration events from the periphery to the CNS
Negative Binomial Models (per transition)
Marker Samples Detectable1 Effect Status exp(Beta) 95% CI p-value Significance Direction
Eotaxin (CCL11) 15 Ok ✅ 1.038 0.932–1.16 0.501 ⬆
GM-CSF 10 Ok ✅ 0.898 0.832–0.97 0.005 ** ⬇
GRO-alpha (CXCL1) 15 Ok ✅ 0.993 0.989–1.00 0.005 ** ⬇
IL-1α 1 Sparse Data ❌ 1.373 0.098–19.24 0.814 ⬆
IL-1β 7 Ok ✅ 0.889 0.765–1.03 0.127 ⬇
IL-1RA 14 Ok ✅ 1.000 1.000–1.00 0.015 * ⬇
IL-2 2 Sparse Data ❌ 0.550 0.437–0.69 0.000 *** ⬇
IL-5 6 Ok ✅ 0.946 0.845–1.06 0.337 ⬇
IL-6 14 Ok ✅ 0.999 0.998–1.00 0.013 * ⬇
IL-7 15 Ok ✅ 0.684 0.217–2.16 0.518 ⬇
IL-8 (CXCL8) 15 Ok ✅ 0.999 0.998–1.00 0.313 ⬇
IL-9 1 Sparse Data ❌ 1.297 0.922–1.82 0.135 ⬆
IL-10 10 Ok ✅ 0.848 0.729–0.99 0.032 * ⬇
IL-15 6 Ok ✅ 0.543 0.325–0.91 0.020 * ⬇
IL-17A/CTLA-8 2 Sparse Data ❌ 0.701 0.295–1.66 0.420 ⬇
IL-18 15 Ok ✅ 0.999 0.986–1.01 0.818 ⬇
IL-31 1 Sparse Data ❌ 0.955 0.787–1.16 0.643 ⬇
IP-10 (CXCL10) 15 Ok ✅ 1.000 0.999–1.00 0.933 ⬆
MCP-1 (CCL2) 15 Ok ✅ 1.000 0.999–1.00 0.851 ⬇
MIP-1α (CCL3) 15 Ok ✅ 0.928 0.852–1.01 0.082 . ⬇
MIP-1β (CCL4) 15 Ok ✅ 0.991 0.974–1.01 0.277 ⬇
RANTES (CCL5) 15 Ok ✅ 1.017 0.965–1.07 0.522 ⬆
SDF-1α 15 Ok ✅ 1.000 1.000–1.00 0.862 ⬆
TNF-α 3 Sparse Data ❌ 0.488 0.357–0.67 0.000 *** ⬇
1 Number of samples with ≥1 measure(s) above the LOD across the 3 replicates.
Associations between CSF biomarkers and counts of migration events across the BBB
Negative Binomial Models (per transition)
Marker Samples Detectable1 Effect Status exp(Beta) 95% CI p-value Significance Direction
Eotaxin (CCL11) 15 Ok ✅ 1.038 0.94–1.15 0.460 ⬆
GM-CSF 10 Ok ✅ 0.909 0.85–0.98 0.008 ** ⬇
GRO-alpha (CXCL1) 15 Ok ✅ 0.994 0.99–1.00 0.006 ** ⬇
IL-1α 1 Sparse Data ❌ 0.881 0.08–9.65 0.917 ⬇
IL-1β 7 Ok ✅ 0.901 0.79–1.03 0.137 ⬇
IL-1RA 14 Ok ✅ 1.000 1.00–1.00 0.005 ** ⬇
IL-2 2 Sparse Data ❌ 0.586 0.47–0.73 0.000 *** ⬇
IL-5 6 Ok ✅ 0.955 0.86–1.06 0.388 ⬇
IL-6 14 Ok ✅ 0.999 1.00–1.00 0.012 * ⬇
IL-7 15 Ok ✅ 0.754 0.26–2.15 0.598 ⬇
IL-8 (CXCL8) 15 Ok ✅ 0.999 1.00–1.00 0.231 ⬇
IL-9 1 Sparse Data ❌ 1.215 0.88–1.67 0.230 ⬆
IL-10 10 Ok ✅ 0.860 0.75–0.99 0.033 * ⬇
IL-15 6 Ok ✅ 0.603 0.37–0.98 0.042 * ⬇
IL-17A/CTLA-8 2 Sparse Data ❌ 0.706 0.32–1.56 0.388 ⬇
IL-18 15 Ok ✅ 0.999 0.99–1.01 0.828 ⬇
IL-31 1 Sparse Data ❌ 0.960 0.81–1.14 0.649 ⬇
IP-10 (CXCL10) 15 Ok ✅ 1.000 1.00–1.00 0.902 ⬆
MCP-1 (CCL2) 15 Ok ✅ 1.000 1.00–1.00 0.649 ⬇
MIP-1α (CCL3) 15 Ok ✅ 0.934 0.86–1.01 0.081 . ⬇
MIP-1β (CCL4) 15 Ok ✅ 0.990 0.98–1.01 0.211 ⬇
RANTES (CCL5) 15 Ok ✅ 1.016 0.97–1.07 0.511 ⬆
SDF-1α 15 Ok ✅ 1.000 1.00–1.00 0.921 ⬆
TNF-α 3 Sparse Data ❌ 0.525 0.40–0.70 0.000 *** ⬇
1 Number of samples with ≥1 measure(s) above the LOD across the 3 replicates.
Associations between CSF biomarkers and counts of migration events within CNS
Negative Binomial Models (per transition)
Marker Samples Detectable1 Effect Status exp(Beta) 95% CI p-value Significance Direction
Eotaxin (CCL11) 15 Ok ✅ 1.304 0.91–1.88 0.154 ⬆
GM-CSF 10 Ok ✅ 0.711 0.35–1.43 0.338 ⬇
GRO-alpha (CXCL1) 15 Ok ✅ 0.519 0.28–0.96 0.038 * ⬇
IL-1α 1 Sparse Data ❌ 0.684 0.52–0.90 0.006 ** ⬇
IL-1β 6 Ok ✅ 0.966 0.53–1.76 0.910 ⬇
IL-1RA 14 Ok ✅ 0.476 0.33–0.69 0.000 *** ⬇
IL-2 2 Sparse Data ❌ 0.576 0.37–0.90 0.015 * ⬇
IL-5 7 Ok ✅ 0.835 0.40–1.75 0.633 ⬇
IL-6 14 Ok ✅ 0.510 0.28–0.92 0.024 * ⬇
IL-7 15 Ok ✅ 1.083 0.67–1.75 0.746 ⬆
IL-8 (CXCL8) 15 Ok ✅ 0.796 0.54–1.17 0.247 ⬇
IL-9 1 Sparse Data ❌ 0.653 0.48–0.89 0.007 ** ⬇
IL-10 9 Ok ✅ 1.061 0.68–1.66 0.797 ⬆
IL-15 7 Ok ✅ 0.666 0.44–1.02 0.060 . ⬇
IL-17A/CTLA-8 2 Sparse Data ❌ 1.232 0.85–1.78 0.266 ⬆
IL-18 15 Ok ✅ 1.492 0.83–2.68 0.181 ⬆
IL-31 1 Sparse Data ❌ 1.146 0.64–2.04 0.644 ⬆
IP-10 (CXCL10) 15 Ok ✅ 1.195 0.86–1.66 0.292 ⬆
MCP-1 (CCL2) 15 Ok ✅ 0.495 0.30–0.82 0.006 ** ⬇
MIP-1α (CCL3) 15 Ok ✅ 0.834 0.46–1.51 0.551 ⬇
MIP-1β (CCL4) 15 Ok ✅ 0.881 0.60–1.30 0.524 ⬇
RANTES (CCL5) 15 Ok ✅ 1.015 0.68–1.51 0.941 ⬆
SDF-1α 15 Ok ✅ 1.079 0.66–1.75 0.758 ⬆
TNF-α 3 Sparse Data ❌ 0.555 0.32–0.95 0.032 * ⬇
1 Number of samples with ≥1 measure(s) above the LOD across the 3 replicates.

AIM 1: Reservoirs Dynamic and Reservoirs Characteristics

Model Results: Longitudinal / state-level data

Viral Suppression: Individuals with detectable viral load >1000 copies/µl are excluded (eligible: LG30, LG44, LG48 and LG51).

Objectives

  • Primary objectives:
    • model viral migration events across the Blood Brain Barrier (CNS ↔︎ periphery).
    • identify reservoir characteristics associated with dynamics
Associations between Reservoirs Measures and counts of migration events from the CNS to the periphery
Negative Binomial Models (per transition)
Reservoir Measures Effect Status exp(Beta) 95% CI p-value Significance Direction
HIV DNA (source) Ok ✅ 0.885 7.1e-01–1.1e+00 0.277 ⬇
HIV DNA (destination) Ok ✅ 1.017 8.6e-01–1.2e+00 0.847 ⬆
Viral Diversity (source) Ok ✅ >1e6 5.4e+04–+Inf 0.006 ** ⬆
Viral Diversity (destination) Ok ✅ >1e6 7.0e-03–+Inf 0.143 ⬆
Viral Divergence Wide CI ❌ 4.141 <1e-6–+Inf 0.904 ⬆
Associations between Reservoirs Measures and counts of migration events from the periphery to the CNS
Negative Binomial Models (per transition)
Reservoir Measures Effect Status exp(Beta) 95% CI p-value Significance Direction
HIV DNA (source) Ok ✅ 1.085 9.8e-01–1.2e+00 0.114 ⬆
HIV DNA (destination) Ok ✅ 1.052 9.6e-01–1.2e+00 0.287 ⬆
Viral Diversity (source) Ok ✅ 82.170 1.8e-04–+Inf 0.507 ⬆
Viral Diversity (destination) Ok ✅ 4.366 3.7e-06–+Inf 0.836 ⬆
Viral Divergence Wide CI ❌ 0.159 <1e-6–+Inf 0.842 ⬇
Associations between Reservoirs Measures and counts of migration events across the BBB
Negative Binomial Models (per transition)
Reservoir Measures Effect Status exp(Beta) 95% CI p-value Significance Direction
HIV DNA (source) Ok ✅ 1.079 1.0e+00–1.2e+00 0.064 . ⬆
HIV DNA (destination) Ok ✅ 0.995 9.3e-01–1.1e+00 0.885 ⬇
Viral Diversity (source) Ok ✅ 950493.238 1.0e+01–+Inf 0.019 * ⬆
Viral Diversity (destination) Ok ✅ 19.716 2.3e-04–+Inf 0.607 ⬆
Viral Divergence Wide CI ❌ 0.254 <1e-6–+Inf 0.863 ⬇
Associations between Reservoirs Measures and counts of migration events within CNS
Negative Binomial Models (per transition)
Reservoir Measures Effect Status exp(Beta) 95% CI p-value Significance Direction
HIV DNA (source) Ok ✅ 0.739 6.3e-01–8.6e-01 0.000 *** ⬇
HIV DNA (destination) Ok ✅ 1.059 9.4e-01–1.2e+00 0.365 ⬆
Viral Diversity (source) Wide CI ❌ 296.502 <1e-6–+Inf 0.571 ⬆
Viral Diversity (destination) Ok ✅ 14999.046 9.3e-04–+Inf 0.256 ⬆
Viral Divergence Wide CI ❌ 5.242 <1e-6–+Inf 0.868 ⬆

AIM 2: CNS Compartmentalization Analysis

Approach

Participants

  • All Last Gift individuals
  • Restricted to virally suppressed donors?
  • Only samples with SGA clones available in both CNS and periphery

1. Data Preparation

  • Multiple sequence alignment (MAFFT)
  • Build maximum-likelihood phylogeny (IQ-TREE)
    • GTR+Γ model and 1,000 ultrafast bootstraps

2. Identify CNS-specific Phylogenetic Clusters

  • Define CNS taxa per participant with custom R functions:
find_cns_clusters(tree, cns_taxa, min_size = 5)
filter_nested_clusters(cluster_list)
  • Scan all internal nodes. A cluster is retained if:
    • ≥ 5 tips - originate from CNS only
  • Nested CNS clades are automatically removed

3. Bayesian Temporal Reconstruction (BEAST)

  • Build time-scaled phylogenies for each participant
  • Model: Bayesian Skyline coalescent (MCMC chains ESS > 200)

4. Outputs

  • Time-scaled phylogeny annotated with CNS clusters
  • Posterior density of tMRCA for each CNS-specific cluster

Workflow

Results Summary

Note

  • Our analyses included 18 participants with sequences from both the CNS and Periphery.
  • Only SGA clones generated from tissue and body fluid samples were kept. (samples from Brain Myeloid cells and T cells will be integrated in future analyses)
🧠 Last Gift cohort HIV Reservoir CNS Clusters
cluster1 Cluster size tMRCA (median) lower HPD upper HPD Basal Ganglia Frontal Cortex (motor) Hippocampus Occipital Cortex Spinal Cord (cervical) Spinal Cord (lumbosacral) Spinal Cord (thoracic) Pons Medulla Parietal Cortex
LG03
Cluster 1 17 2005-02-01 1994-04-01 2011-05-22 4 (24%) 1 (6%) 2 (12%) 1 (6%) 3 (18%) 3 (18%) 3 (18%) – – –
LG05
Cluster 1 17 2015-09-11 2013-05-12 2017-03-06 6 (35%) 5 (29%) 1 (6%) 1 (6%) – – – 4 (24%) – –
LG14
Cluster 1 21 2014-06-10 2011-12-16 2016-05-11 7 (33%) – – – 5 (24%) 4 (19%) – 5 (24%) – –
LG18
Cluster 1 6 2005-12-08 1994-11-26 2014-03-02 – – – – – – – – 6 (100%) –
LG22
Cluster 1 10 2022-05-26 2022-01-14 2022-08-28 – – – – – – 8 (80%) – 2 (20%) –
LG23
Cluster 1 6 2020-10-29 2020-06-15 2021-02-12 – – – 1 (17%) 1 (17%) 4 (67%) – – – –
LG25
Cluster 1 8 2020-11-27 2020-08-15 2021-02-09 4 (50%) 2 (25%) – – – – 2 (25%) – – –
LG29
Cluster 1 6 2021-01-27 2019-07-31 2021-07-10 6 (100%) – – – – – – – – –
LG30
Cluster 1 20 2020-03-27 2018-02-20 2021-07-13 8 (40%) – 1 (5%) – – – 1 (5%) 8 (40%) 2 (10%) –
Cluster 2 7 2018-05-23 2012-03-28 2021-03-05 – – 5 (71%) 2 (29%) – – – – – –
Cluster 3 7 2021-10-31 2020-04-27 2022-09-08 – – – – – – – – – 7 (100%)
LG33
Cluster 1 38 2008-09-26 2003-03-23 2012-11-15 5 (13%) 8 (21%) 1 (3%) 7 (18%) – 5 (13%) – 3 (8%) – 9 (24%)
Cluster 2 15 2008-06-27 2002-06-12 2012-10-19 – – – – 4 (27%) – 8 (53%) – 3 (20%) –
LG42
Cluster 1 22 2023-05-14 2023-02-03 2023-07-29 – – – – 3 (14%) 5 (23%) 6 (27%) 7 (32%) 1 (5%) –
LG44
Cluster 1 12 2021-06-27 2019-11-11 2023-03-15 – – – – 1 (8%) – 4 (33%) 3 (25%) 4 (33%) –
Cluster 2 17 2020-01-31 2017-11-24 2021-06-29 – – – – 6 (35%) 5 (29%) 2 (12%) 2 (12%) 2 (12%) –
LG48
Cluster 1 6 2024-02-18 2023-02-24 2024-09-29 – – – – 1 (17%) 2 (33%) 1 (17%) 1 (17%) 1 (17%) –
Cluster 2 7 2023-07-27 2021-07-20 2024-08-02 – – – – – – – 3 (43%) 4 (57%) –
LG49
Cluster 1 8 2021-07-09 2017-06-11 2023-08-19 – – – – 4 (50%) – 2 (25%) 2 (25%) – –
1 Using most recently single genome sequencing env data

Phylogenies

Interpretation/Discussion

  • Early vs late CNS seeding
  • Evidence for CNS-restricted evolution
  • Within-host compartmentalization patterns

  • Inclusion Criteria:

    • virally suppressed
    • SGA available in both CNS and periphery
    • CSF data

it would end up with a population size of 16 :

final_population
 [1] "LG03" "LG05" "LG12" "LG14" "LG15" "LG18" "LG22" "LG24" "LG28" "LG29"
[11] "LG33" "LG34" "LG37" "LG39" "LG43" "LG49"

Thank you